Rapid Phenotyping of Knockout Mice to Identify Genetic Determinants of Bone Strength

231 1

b freudenthal and others Rapid skeletal phenotyping of 231:1 R31–R46 Review knockout mice

Open Access

Rapid phenotyping of knockout mice to identify genetic determinants of bone strength

Correspondence should be addressed Bernard Freudenthal1, John Logan1, Sanger Institute Mouse Pipelines2, to J H D Bassett or Peter I Croucher3, Graham R Williams1 and J H Duncan Bassett1 G R Williams Email 1Molecular Endocrinology Laboratory, Department of Medicine, Imperial College London, London, UK [email protected] or 2Mouse Pipelines, Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, UK graham.williams@imperial. 3Garvan Institute of Medical Research, Sydney, New South Wales, Australia ac.uk

Abstract

The genetic determinants of osteoporosis remain poorly understood, and there is a Key Words large unmet need for new treatments in our ageing society. Thus, new approaches ff osteoporosis for gene discovery in skeletal disease are required to complement the current ff bone genome-wide association studies in human populations. The International Knockout ff genetics Mouse Consortium (IKMC) and the International Mouse Phenotyping Consortium (IMPC) ff gene discovery provide such an opportunity. The IKMC generates knockout mice representing each of Endocrinology

of the known protein-coding genes in C57BL/6 mice and, as part of the IMPC initiative, the Origins of Bone and Cartilage Disease project identifies mutants with significant outlier skeletal phenotypes. This initiative will add value to data from large human cohorts and Journal provide a new understanding of bone and cartilage pathophysiology, ultimately leading to the identification of novel drug targets for the treatment of skeletal disease. Journal of Endocrinology (2016) 231, R31–R46

Introduction

A novel strategy for osteoporosis gene discovery whose aim is to disrupt each of the protein-coding genes in C57BL/6 mice, and the International Mouse Studies of human monogenic extreme phenotype Phenotyping Consortium (IMPC) that has established disorders have been instrumental in discovering a multidisciplinary and broad primary phenotype genetic and molecular mechanisms of common screen to characterise these mutant mice. By using diseases including obesity and diabetes (Yamagata et al. 1996, Montague et al. 1997). However, collection of samples from mice that have undergone the IMPC human extreme phenotype cohorts takes many years phenotyping pipeline, a bespoke rapid-throughput and requires significant effort and financial resource. multi-parameter skeletal phenotyping platform has A new approach to osteoporosis gene discovery been applied systematically to detect significant involves systematic identification of extreme skeletal phenotypes by screening minimal number of samples. phenotypes in mutant mouse lines that carry single- This phenotyping programme exploits the excellent gene knockouts representing all the known protein- replication of human skeletal disease in mice, and novel coding genes. This approach has been made possible by susceptibility genes can be validated by interrogating the International Knockout Mouse Consortium (IKMC), human osteoporosis cohorts.

http://joe.endocrinology-journals.org © 2016 The authors This work is licensed under a Creative Commons DOI: 10.1530/JOE-16-0258 Published by Bioscientifica Ltd. Attribution 3.0 Unported License. Printed in Great Britain Downloaded from Bioscientifica.com at 09/29/2021 03:45:34AM via free access

10.1530/JOE-16-0258 Review b freudenthal and others Rapid skeletal phenotyping of 231:1 R32 knockout mice

‘Known unknowns’ in osteoporosis regulate the function of osteoblasts and osteoclasts. Both these pathways have subsequently been targeted by novel Osteoporosis is a worldwide healthcare problem that osteoporosis treatments. causes up to 9 million fractures annually (Johnell & Kanis The canonical Wnt/β-catenin signalling pathway 2006). Within the EU, it is estimated that osteoporosis is the key regulator of osteoblasts, which mediate bone affects 30 million people and osteoporotic fractures formation (Balemans et al. 2001, Gong et al. 2001, Boyden cost €37 billion annually (Hernlund et al. 2013). These et al. 2002, Little et al. 2002, Glass et al. 2005). The major numbers are projected to rise with the increasing elderly Wnt antagonist sclerostin (SOST) was first identified by population. Osteoporotic hip fractures are associated studying subjects with high bone mass due to sclerosteosis with a significant rise in mortality Sattui( & Saag 2014), and Van Buchem disease (Balemans et al. 2001, Brunkow especially during the year following fracture when it is et al. 2001). The importance of Wnt signalling was further estimated to be 8–36% (Abrahamsen et al. 2009). highlighted by the discovery of activating and inactivating The most important risk factors for osteoporotic mutations of the Wnt co-receptor LRP5, which results fracture are low bone mineral density (BMD) (clinically in high and low bone mass, respectively (Johnson et al. assessed by dual-energy X-ray absorptiometry (DEXA or 1997, Little et al. 2002). DXA)), increasing age and history of fracture (Johnell The RANK-RANKL-OPG pathway regulates osteoclasts, et al. 2005). There are two key determinants of adult which mediate bone resorption. This signalling pathway BMD: the peak bone mass attained in early adulthood was discovered in a functional screen of tumour necrosis and the rate of bone loss during ageing. Variation in BMD factor (TNF)/TNF receptor superfamily members. OPG has a large heritable genetic component. This is known is an endogenous inhibiting decoy receptor of RANKL from observations of familial clustering of osteoporosis related to the TNF receptor. In vivo overexpression of (Seeman et al. 1989, Keen et al. 1999) and from twin studies OPG in mice was found to cause osteopetrosis due to that have calculated that the heritable contribution lies impairment of the later stages of osteoclast differentiation between 40 and 90% (Mitchell & Yerges-Armstrong 2011). (Simonet et al. 1997). Subsequently, human osteopetrosis The heritable contribution to variance in BMD is greatest

Endocrinology phenotypes with increased BMD were found to be caused in early adulthood (Gueguen et al. 1995), yet variation in of by mutations of RANK, RANKL and other related genes the rate of bone loss per se is also genetically determined involved in osteoclast differentiation (Coudert et al. 2015). (Kelly et al. 1993). Thus, genetic mechanisms contribute

Journal significantly to the risk of osteoporosis. Nevertheless, the known BMD-associated genetic variants account for only Regulatory mechanisms in bone turnover 5.8% of the total variance (Estrada et al. 2012), indicating Knowledge of the signalling pathways that regulate that the majority of susceptibility genes have yet to bone turnover is essential for understanding the be identified. pathophysiology of osteoporosis. The manifest complexity of the signalling pathways and networks that regulate the cellular processes involved in dynamic bone turnover is Gene discovery from skeletal extreme phenotypes significant as small differences in function of individual Intrinsic and extrinsic factors, systemic hormones, components, including those not yet discovered, may neuronal innovation and mineral homeostasis can all have a combined effect on heritable risk of osteoporosis. affect bone mass. Significantly, many of the genes and The opposing processes of bone resorption and formation signalling pathways involved in the intrinsic regulation are tightly regulated by critical mechanisms including the of bone turnover and bone mass have been identified Wnt signalling and RANKL/RANK/OPG pathways (Fig. 1). by the study of human monogenic disorders associated At the cellular level, bone remodelling takes place in with extremes of BMD (Table 1). In traditional gene multicellular units, which comprise co-located osteoclasts discovery, the loci of causative alleles would be identified and osteoblasts within a bone remodelling cavity (Raggatt by linkage analysis in the families of index cases, followed & Partridge 2010). The bone remodelling cycle is initiated by positional cloning of the relevant genes (Alonso & by osteocytes in response to altered mechanical loading Ralston 2014). Such studies have identified the two (Nakashima et al. 2011), local microdamage and systemic key regulatory pathways – canonical Wnt signalling factors such as parathyroid hormone (Goldring 2015). and receptor activator of nuclear factor kappa-B ligand Unloading stimulates expression of RANKL and the Wnt (RANKL)/RANK/osteoprotegerin (OPG) that respectively inhibitors sclerostin and Dickkopf-related protein 1 (DKK-1)

http://joe.endocrinology-journals.org © 2016 Society for Endocrinology Published by Bioscientifica Ltd. DOI: 10.1530/JOE-16-0258 Printed in Great Britain Downloaded from Bioscientifica.com at 09/29/2021 03:45:34AM via free access Review b freudenthal and others Rapid skeletal phenotyping of 231:1 R33 knockout mice

Table 1 Monogenic disorders that have identified key skeletal genes in bone remodelling.

Disease Clinical features Gene Mechanism Reference Reduced Osteoporosis- Reduced bone mass LRP5 Loss-of-function (Gong et al. 2001) bone mass pseudoglioma and blindness mutations disrupt Wnt syndrome (OPPGS) signalling and reduce osteoblastic bone formation. Osteogenesis Increased bone COL1A1, COL1A2, Loss-of-function (Baldridge et al. 2008; imperfecta fragility; blue sclerae CRTAP, LEPRE, mutations in collagen Sykes et al. 1986; in some types PPIB and collagen- van Dijk et al. 2009) processing proteins cause abnormal osteoid matrix, thereby impairing normal bone formation. Juvenile-onset Paget Short stature, TNFRSF11B (OPG) Loss-of-function (Chong et al. 2003) disease fractures, skull mutations disrupt enlargement, inhibition of RANKL by progressive deafness osteoprotegerin, causing increased osteoclastic resorption of bone. X-linked osteoporosis Juvenile-onset PLS3 Loss-of-function (van Dijk et al. 2013) fractures in males mutations affect Plastin-3, an actin- binding protein. Mechanism osteoporosis is unknown. Increased Osteopetrosis Increased bone mass TNFRSF11A Loss-of-function (Frattini et al. 2000; bone mass with fractures (RANK), TNFSF11 mutations affecting Guerrini et al. 2008; Endocrinology (RANKL), CLCN7, osteoclast Kornak et al. 2001; of TCIRG1, OSTM1 differentiation and Pangrazio et al. 2006; function cause reduced Sobacchi et al. 2007) bone resorption. Journal Sclerosteosis and Increased bone mass, SOST Loss-of-function (Balemans et al. 2001) Van Buchem disease syndactyly, entrap- mutations affect ment neuropathies inhibition of Wnt signalling by sclerostin, causing increased osteoblastic bone formation. Autosomal dominant Increased bone LRP5 Gain-of-function (Boyden et al. 2002) high bone mass density, entrapment mutation in Wnt neuropathies, co-receptor causes square jaw and increased osteoblastic torus palatinus bone formation.

in osteocytes, thus increasing osteoclastic bone resorption osteoblastic bone formation by secreting sclerostin and decreasing osteoblastic bone formation. Repetitive (Heino et al. 2002, Tu et al. 2012). If TGFβ levels fall, loading can cause fatigue-induced microdamage and focal bone-lining cells become activated and join osteocytes in tissue injury that leads to osteocyte apoptosis, following secreting the cytokines monocyte/macrophage colony- which pro-osteoclastogenic signalling is initiated by a stimulating factor (M-CSF) and RANKL that stimulate discrete population of adjacent osteocytes (Kennedy et al. recruitment and differentiation of circulating osteoclast 2012). By contrast, increased mechanical loading decreases monocyte progenitors (Nakashima et al. 2011). Mature expression of sclerostin and DKK-1 in osteocytes and leads multi-nucleated bone-resorbing osteoclasts adhere to to increased osteoblastic bone formation (Robling et al. the cell surface and create resorption pits in the middle 2008). In their basal state, osteocytes maintain quiescence of the multicellular unit. The osteoclasts secrete acid and by inhibiting osteoclastogenesis by secreting transforming proteases including cathepsin-K that degrade the bone growth factor β (TGFβ) and inhibiting Wnt-activated matrix, leaving demineralised collagen that is resorbed

Figure 1 Schematic representing the bone remodelling processes of the ‘basic multicellular unit’ in the endosteal surface of trabecular bone. Activation: microdamage to the bone causes osteocyte apoptosis, reducing local basal inhibition of osteoclastogenesis. Resorption: in response to PTH signalling, RANKL and CSF-1 (colony-stimulating factor-1) increase recruitment, proliferation and differentiation of osteoclasts, which demineralise the bone matrix and then digest the collagen matrix, the remnants being removed by macrophages. Formation: PTH and mechanical activation of osteocytes reduce sclerostin expression, removing the potent inhibition of Wnt-mediated osteoblast differentiation (via cell surface receptor Frizzled and co-receptors LRP5 and LRP6) and bone formation. Termination: in response to increasing levels of sclerostin, bone formation ceases and newly deposited osteoid is Endocrinology mineralised. of

by macrophages. Wnt-activated osteoblasts subsequently New treatments for osteoporosis include small- Journal secrete and mineralise new bone matrix (osteoid) to molecule inhibitors and biological therapies have been fill the resorption cavity. When the repair is complete, developed based on this new understanding of bone the bone surface returns to its quiescent state and turnover. Denosumab is a fully humanised monoclonal sclerostin stimulates mineralisation of the new osteoid antibody to RANKL that mimics the endogenous (Tu et al. 2012). inhibiting activity of OPG (Lacey et al. 2012). Denosumab is administered by 6-month injection; yet, its use is restricted by high cost and its effects are rapidly reversible. Gene discovery drives therapeutic innovation Another new class of osteoporosis drugs are those that Identification of the genes and regulatory networks that target the osteoclast-secreted protease cathepsin-K determine normal bone formation, maintenance and (Makras et al. 2015). However, most of the cathepsin-K strength has facilitated the recent development of new inhibitors have adverse off-target effects, with the targeted treatments to increase BMD and reduce fracture exception of odanacatib, which has now successfully risk. There is a pressing need for new treatment options completed phase III trials (Bone et al. 2015). Humanised in osteoporosis as current treatments achieve at best monoclonal antibodies to the Wnt antagonist sclerostin partial relative risk reduction of fracture. Bisphosphonates (romosozumab and blosozumab) are in development as remain the mainstay of treatment for primary and new anabolic agents (Shah et al. 2015, Sim & Ebeling secondary prevention of osteoporotic fracture, despite 2015). However, since romosozumab is currently in phase the recognition uncommon but significant side effects III trials and blosozumab has only completed phase II (Kennel & Drake 2009). Other treatments including trials, effect on fracture risk is still unknown. strontium ranelate and raloxifene are no longer advocated Despite some recent advances, there remains as first line because of associated cardiovascular risks a large unmet need for new therapeutic targets in (Barrett-Connor et al. 2006, Abrahamsen et al. 2014). osteoporosis. Of all the newer targeted treatments, only

the small-molecule odanacatib is administered orally. The inheritance of the SNPs or CNVs with disease phenotypes requirement for parenteral subcutaneous administration (Manolio 2010). is an impediment to use of teriparatide and all monoclonal To date, GWAS have identified more than 60 loci antibody treatments. It is hoped that further advances in associated with BMD, confirming the polygenic nature knowledge of the complex mechanisms of bone turnover of this variable (Richards et al. 2012). Less data are, regulation and the genetic basis of the heritability of however, available for fracture risk than BMD (Mitchell osteoporosis will enable the accelerated development & Streeten 2013). Although estimation of areal BMD of new osteoporosis treatments. The current goals in by DEXA scanning is a good predictor of fracture osteoporosis gene discovery must be to identify novel susceptibility, it does not necessarily reflect bone quality targets with a systematic methodology that may replicate as DEXA does not account for bone geometry, size and the successes of past studies in Mendelian extreme strength (Marshall et al. 1996). phenotype cohorts. Population-based genetic cohort studies have significant limitations in relation to osteoporosis gene discovery. When GWAS data are combined in meta- Systematic osteoporosis gene discovery analyses, variability in the characterisation of phenotype in human studies can be an important confounder, and population studies are also limited by the quality and reproducibility of the The Human Genome Project and association studies phenotyping (Visscher et al. 2012). Many of the genomic In order to identify novel genes that relate to heritability loci identified by GWAS map to genes in pathways of osteoporosis systematically, new population-based already known to be related to bone biology, such as the methodologies were required and have been used Wnt/β-catenin, RANK-RANKL-OPG, mesenchymal stem extensively. Linkage analysis and positional cloning cell differentiation and SOX9-regulated endochondral techniques, used to identify gene mutations in ossification pathways Mitchell( & Streeten 2013). Overall, Mendelian disorders, were less successful when applied GWAS analysis has not yet led to the discovery of major

Endocrinology to population cohorts with low BMD. Several genome- new mechanisms that regulate bone turnover regulation of wide linkage studies were performed on high and low and predispose to osteoporosis. Indeed, contrary to BMD cohorts to detect quantitative trait loci (QTLs) initial expectations, GWAS analyses failed to discover that regulate BMD. However, these studies produced sufficient determinants of polygenic heritable traits to Journal heterogeneous results, and when combined in a meta- enable individual disease risk prediction. An osteoporosis analysis, none of the QTLs reached genome-wide risk prediction model using the combined weighted significance Ioannidis( et al. 2007). effects of 63 BMD-decreasing alleles in a population- It was the subsequent development of GWAS analysis based study containing 2836 women explained only that made it possible to search for genetic influences on 5.8% of the total variance in femoral neck BMD complex polygenic traits such as osteoporosis. After the (Estrada et al. 2012). first sequence of the human genome was completed, A further difficulty in identifying susceptibility genes the International HapMap Project was established to by GWAS analysis arises from the fact that identified SNPs catalogue the common genetic variation between the are rarely functionally significant, although their loci may genomes of 249 individuals in four different populations, identify adjacent pathogenic pathways and mechanisms. leading to the mapping of over 1 million single-nucleotide The co-association of the SNP with the disease phenotype polymorphisms (SNPs) within the human genome indicates that it is in linkage disequilibrium with a gene (or (International HapMap Consortium 2005). A major a non-coding regulatory element) that has a constituent aim of the HapMap Project was to facilitate the study functional role that is contributory to the phenotype of the contribution of normal genetic variation to the (Cantor et al. 2010). The size of the detected effect is itself inheritance of complex polygenic phenotypic traits such of little importance, and it is typical that the effect size as osteoporosis (Frazer et al. 2009). In GWAS analyses, large of individual identified loci is inversely proportional to population-based cohorts containing many thousands the population sample size in a GWAS. The mean effect of subjects are genotyped by chip-based microarrays for size of SNPs identified in the largest osteoporosis GWAS the presence of millions of SNPs, or alternatively copy- meta-analysis to date was 0.048 standard deviations (s.d.s) number variants (CNVs). These data are used to perform and the largest effect size was 0.1 s.d. (Estrada et al. 2012, hypothesis-free significance testing for association of the Zheng et al. 2015).

To address the limitation that standard GWAS microarrays (Wu et al. 2013). Future work will increasingly microarrays only test common genetic variants (minor use next-generation sequencing methods, which do not allele frequency (MAF) > 5%), larger reference panels have require a priori knowledge of the transcriptome sequence been created from the recent UK10K and 1000 Genomes and have a superior sensitivity, specificity and dynamic projects. ‘Next-generation sequencing’ techniques have range in comparison with current microarrays (Vikman enabled direct imputation of low-frequency coding and et al. 2014). With use of advanced statistical techniques, it non-coding variants found by whole-genome sequencing is possible to discover new splice variants and non-coding of large osteoporosis cohorts. Thus far, only one low- sequences. Despite these advances, RNA-sequencing frequency variant with genome-wide significance for approaches require preparation of high-quality RNA from skeletal disease has been identified by these techniques sufficient numbers of physiologically relevant samples. (Zheng et al. 2015). The variant is near a novel locus, Accordingly, it is not yet possible to study gene expression engrailed homeobox-1 (EN1) and has an estimated effect in osteoporosis using large population-based sample sizes, size on lumbar spine BMD of 0.2 s.d.s, which is four-fold and RNA sequencing is likely to be most useful in the larger than the mean effect size of previously reported analysis of animal models. common variants.

Proteomics Transcriptomics and osteoporosis Proteomics allows the simultaneous analysis of One limitation of the GWAS data is that the majority of all proteins in a sample of cells by antibody-based identified SNPs may be located within poorly annotated purification and mass spectrometry techniques. A number regulatory elements in intronic or intergenic non- of studies have demonstrated differential regulation of coding regions (Cooper 2010). Indeed, it is increasingly cellular protein expression in osteoporosis (Wu et al. appreciated that heritable variation of osteoporosis risk is 2013). Most of the studies have been performed in vitro manifest in more ways than just variation in the coding using osteoblast and osteoclast cell cultures. Human sequence of genes. Expression levels of RNA and proteins, Endocrinology proteomic studies have predominantly used peripheral of as regulated by gene regulation networks and epigenetic circulating monocytes as precursors to osteoclasts (Deng control, are also of great importance. Consequently, et al. 2011a,b), or bone marrow-derived mesenchymal study of variations in levels of RNA transcripts has also Journal stem cells (Choi et al. 2010). However, reproducibility of been applied to identify new molecular mechanisms such results has been limited. and signalling pathways involved in regulation of BMD. Furthermore, it is anticipated that analysis of non-coding RNAs will lead to a better understanding Epigenomics of the significance of intergenic loci identified by GWAS (Hangauer et al. 2013). The heritability of osteoporosis may also be mediated Transcriptomic studies have used high-throughput by epigenetic gene regulation, in which gene or allele microarray analysis to quantify mRNA expression expression is ‘imprinted’ by DNA methylation and in osteoporotic and non-osteoporotic human bone, histone modifications Holroyd( et al. 2012). Epigenome- precursor mesenchymal stem cells and primary osteoblast wide studies have only recently been performed in cultures (Wu et al. 2013). In a microarray study of gene relation to osteoporosis. Transcriptome gene expression expression in human osteoblasts, 1606 genes were found microarray, epigenomic miRNA microarray and to be differentially expressed between osteoporotic and methylome sequencing were simultaneously performed non-osteoporotic subjects (Trost et al. 2010). Many of these using circulating monocytes from five subjects with genes had been identified previously by GWAS analysis, low hip BMD and five with high BMD, in order to but several were new and could not have been predicted. integrate transcriptomic and epigenomic data (Zhang Thus, osteoporosis transcriptomics studies have confirmed et al. 2015). The aim is to reveal the higher regulatory that numerous distinct signalling pathways are involved mechanisms (‘interaction network modules’) underlying in the regulation of bone turnover and bone mass. the genetic control of osteoporosis heritability. However, However, replication of these studies has been difficult and the problems of low sample number and difficulty confounded by limited sample size, poor signal to noise of accessing human skeletal tissues are important ratio and technical limitations of commercially available limitations in such approaches.

The mouse as an essential tool for cryo-preserve mouse lines for unrestricted use by the osteoporosis gene discovery research community. The IKMC was established with the goal of creating a complete resource of reporter-tagged Human population studies have been limited in their ability null mutations for all protein-coding genes in C57BL/6 to identify novel genetic determinants of osteoporosis, mouse ES cells by 2021 (Collins et al. 2007). although they have confirmed that bone-regulating The IKMC-targeted mutagenesis techniques involve use genes identified in the study of rare extreme phenotypes of bacterial artificial chromosome (BAC)-based ‘Knockout- contribute to the complex heritability of secondary First’ conditional ready gene targeting vectors that use osteoporosis. Across the breadth of human biology, there homologous recombination to incorporate a lacZ selection are huge knowledge gaps as the function of most genes and cassette in the relevant targeted sequence (Fig. 2) (Testa the heritability of many complex diseases remain unknown et al. 2004). This enables the generated mouse lines to be (Schofield et al. 2012). In order to meet the challenge versatile and powerful tools for research. The Knockout- of discovering the functional relationship between First vectors combine two functions by creating either a human genes and their phenotypic effects, powerful new reporter-tagged knockout or a conditional mutation if the experimental tools are required. Model organisms provide gene-trap cassette is removed by FLP recombinase, thereby the ability to draw genetic and physiological parallels to reverting the knockout mutation to wild type, although human genetic systems, and their use has complemented with addition of loxP sites that flank a functionally critical many advances in human molecular genetics. In particular, exon (Skarnes et al. 2011). The consequence of this is that an extensive range of techniques for experimental genetic temporal or tissue-specific analysis of gene function can be manipulation have been developed in mice. performed by crossing mice bearing the allele containing The genome of the C57BL/6J mouse strain was the LoxP-flanked (‘floxed’) critical exon with transgenic sequenced in 2002 (Waterston et al. 2002), shortly after mice that express Cre recombinase under control of a the human genome. Thus, laboratory mice provide a constitutive or inducible cell type-specific promoter. unique resource that can be used to generate genetically modified models of human diseases Schofield( et al. Endocrinology Pipeline phenotyping of knockout mice of 2012). Currently, 17,055 mouse–human homologs have been annotated in the Mouse Genome Database (Dolan The IKMC provides readily available knockout mice that et al. 2015). Critically, the selective mutation or deletion Journal can be used for functional investigation of candidate (‘knockout’) of individual genes in mice can be used to genes, for example in loci identified by GWAS analysis identify and characterise gene function in vivo. (Cox & Church 2011). The singular ambition to deduce the function of all the genes discovered in the Human Genome Project, however, has led to the development of International knockout mouse programmes ambitious large-scale systematic phenotyping programmes Such is the utility of knockout mice that a number of all knockout mouse lines with deletions of homologs to of mutagenesis programmes have collaborated to human genes (Bradley et al. 2012). These projects enable

Figure 2 Knockout-first strategy for creating dual-purpose knockout/conditional alleles. Bacterial artificial chromosome (BAC)-based targeting vectors are inserted by homologous recombination into mouse ES cells. Recombination steps with Cre or Flp recombinase are illustrated. (A) Knockout-first allele (reporter-tagged insertion allele). Gene-trap knockout is generated using a targeting cassette containing the marker genes lacZ and neomycin. A separate loxP site is inserted on the other side of a critical exon (Exon 2). (B) Conditional allele (post-Flp). By crossing mice with a Flp deleter strain, the gene-trap knockout is reversed and a floxed allele is created, enabling conditional Cre recombinase-mediated gene inactivation. (C) Deletion allele (post-Flp and Cre with no reporter).

the study of the extreme phenotype potential of every phenotyping are body length and X-ray skeletal survey to known coding gene, thereby providing the potential detect gross anatomical variation. As critical determinants for systematic discovery of novel critical skeletal genes. of healthy bone include bone mineral content (BMC), What would otherwise have been individually impossible bone strength and other three-dimensional morphological in terms of resources and complexity has been achieved parameters, the IMPC phenotype screen lacks the by the coordination and the standardisation of a series precision required for comprehensive detection of bone of international programmes to form a major worldwide structure and strength abnormalities. project The IMPC (Koscielny et al. 2014). To address these limitations, a number of ancillary The international phenotyping pipeline incorporates specialist screens have been established. The Origins of standardised and validated tests with protocols shared by Bone and Cartilage Disease (OBCD) project (http://www. contributing mouse clinics across the world (de Angelis boneandcartilage.com/) is performing a multi-parameter et al. 2015). The pipelines incorporate 20 phenotyping tests skeletal phenotyping screen of mouse lines generated by that capture 413 parameters and cover all systems divided the WTSI, an IMPC partner institution, using methods between the categories of morphology, metabolism, that give critical functional and structural information cardiovascular, bone, neurobehavioural and sensory, regarding skeletal phenotype (Fig. 3). The goals of the haematology, clinical chemistry and allergy/immune. The OBCD project are to (i) identify novel pathways regulating phenotyping tests are performed at defined times in the normal bone development, maintenance and resilience; first 16 postnatal weeks and require a minimum cohort (ii) uncover new genetic determinants of osteoporosis; of seven male and seven female mice (Brown & Moore and (iii) provide in vivo models to elucidate their molecular 2012). New statistical models have had to be developed, basis and investigate novel treatments. in order to ensure the reproducibility of the multivariate New imaging and biomechanical techniques have data generated (Karp et al. 2015). In an initial report of been developed to detect abnormalities of bone structure phenotype data on mutant lines representing 320 unique and strength that parallel those occurring in human genes, 83% of mutant lines were outliers (de Angelis et al. disease. Cross-disciplinary collaboration with the fields

Endocrinology 2015). Of the 250 lines reported in an early phase of the of biophysics, microimaging and statistics has enabled of phenotyping pipeline contributed by the IMPC member development of a bespoke rapid-throughput multi- Wellcome Trust Sanger Institute (WTSI), 104 were lethal parameter bone phenotyping platform (Fig. 4) (Bassett or sub-viable and were phenotyped as heterozygotes; et al. 2012a, Esapa et al. 2012). Journal nonetheless, haploinsufficient phenotypes were detected in 38 of these lines (White et al. 2013). Ongoing results from the international multicentre phenotyping projects OBCD phenotyping techniques are released at http://www.Mousephenotype.org/. Limbs and caudal vertebrae from knockout mice are analysed at 16 weeks of age after completion of the IMPC phenotyping pipeline at WTSI. No additional animals Skeletal phenotyping screen of knockout are required for the OBCD screen, and transportation of mouse lines samples is logistically simple. Each batch of 75 samples, which contains both mutants and wild-type controls, is Ancillary IMPC projects analysed blind with genotypes only being assigned on The systematic phenotyping of knockout mice by the completion of the batch’s phenotyping. IMPC has already begun to reveal genes whose deletion Digital X-ray microradiography is performed on has an effect on BMD (estimated by whole-body DEXA), femurs and tail vertebrae, and the images are analysed to implying a functional role in bone development and determine bone length and BMC (Bassett et al. 2012b). skeletal physiology. To date, 79 out of 1820 lines have an X-ray images are recorded at 10 µm pixel resolution using a outlier phenotype associated with decreased BMD (IMPC Faxitron MX20 specimen radiography system. Bone length phenotype MP:0000063) and 41 lines have increased is determined using ImageJ 1.41 software (http://rsb.info. BMD (IMPC phenotype MP:0000062). However, a major nih.gov/ij/). The relative mineral content of the calcified limitation is the poor sensitivity and specificity of DXA tissues is quantified after calibrating each image to three for the analysis of the mouse skeleton (Holmen et al. internal standards. Measurement of the median grey level 2004). Other skeletal parameters included in the IMPC is used to identify outliers with increased or decreased BMC

Journal Figure 3 Flow chart showing how the OBCD bone phenotyping platform leads to identification of significant abnormal skeletal phenotypes, in conjunction with the IMPC standardised phenotyping project.

relative to reference data obtained for more than 100 wild- plotted so that yield load, maximum load and fracture load type controls of the same genetic background. can be determined. Stiffness is determined from the slope Micro-CT analysis has been optimised to determine of the linear (elastic) part of the load–displacement curve. femoral trabecular bone parameters including trabecular bone volume as proportion of tissue volume (BV/TV), Pilot study trabecular thickness and trabecular number. In addition, cortical bone parameters of cortical thickness, cortical Before commencing the OBCD skeletal phenotyping diameter and cortical volumetric BMD are calculated screen, a prospective pilot study of 100 unselected (Bouxsein et al. 2010, Esapa et al. 2012). A Scanco knockout mouse lines was undertaken (Bassett et al. micro-CT 50 system allows automated rapid-throughput 2012a). As part of the pilot project, it was necessary to high-resolution imaging. Three-dimensional quantitative determine reference ranges and coefficients of variation image analysis is performed using clearly defined regions for each of the study parameters for female C57BL/6 wild- of interest and compared with reference data. type mice. Principal component analysis was used to Biomechanical variables of bone strength and optimise the detection of significant skeletal phenotypes toughness are derived from femur three-point bend test in multivariate outliers, as significant abnormal pheno load–displacement curves and tail vertebrae compression types may only be detected when variances in all para testing (Esapa et al. 2012). Load–displacement curves are meters are considered (Rousseeuw & Van Zomeren 1990).

Endocrinology Figure 4 of OBCD skeletal phenotyping methods. (A) X-ray microradiography images of femur and fifth to sixth tail vertebrae from wild-type and mutant mice. Low bone mineral content is represented in green/yellow colour and high bone mineral content is represented in red/pink colour in pseudo-coloured images. Cumulative frequency graphs showing difference in bone mineral content between wild-type and mutant mice. (B) Micro-CT images of cortical and

Journal trabecular bone from wild-type and mutant mice. Cortical thickness and trabecular bone volume/total volume (BV/TV) parameters in the mutant are shown in comparison with reference mean ± 2 standard deviations. (C) Femur three-point bend and vertebral compression analysis with load–displacement curves illustrating biomechanical parameters.

As reference data were established in a large number of included five genes whose deletion results in low bone genetically identical wild-type mice, power calculations mass (Bbx, Cadm1, Fam73b, Prpsap2 and Slc38a10) and determined that samples from only two animals were four whose deletion results in high bone mass (Asxl1, required to detect a significant abnormality that represents Setdb1, Spns2 and Trim45). The study also confirmed the an outlier phenotype. low bone mass phenotype identified previously inSparc In the pilot study of 100 knockout mouse lines, nine knockout mice (Delany et al. 2000). Three of the knockout new genetic determinants of bone mass and strength were lines carried heterozygous mutations (Asxl1, Setdb1 and identified, none of which had been identified previously Trim45), whereas the rest were homozygotes. in GWAS analysis of osteoporosis cohorts or could have been predicted a priori. Analysis of the contrasting Other skeletal phenotyping programmes patterns of abnormality amongst the different phenotypic parameters enabled phenotypes to be categorised in a way Although the OBCD pilot study was the first approach to that could be mapped directly to human skeletal disease. be published, similar phenotype screening methods have Bones were classified as either (i) weak but flexible with been undertaken by others. Lexicon Pharmaceuticals, low BMC (as in osteoporosis), (ii) weak and brittle with low Inc. recently published selected results from a screen of BMC (typical of matrix disorders including osteogenesis knockout mouse lines to search for potential osteoporosis imperfecta) or (iii) strong but brittle with high BMC (high drug targets (Brommage et al. 2014). This phenotyping bone mass disorders such as osteopetrosis). The nine screen included three techniques (skeletal DEXA of new determinants of bone mass and strength identified live mice, micro-CT of dissected bones and histological

examination of decalcified bones). Ten novel genes were selection procedure follows a specific algorithm Fig. 5( ). named, and three further unnamed novel genes coding Although novelty is a key criterion, phenotype severity, for apparent potential osteoporosis drug targets were biological plausibility, human disease association and alluded to. experimental tractability are also critical considerations The IMPC-constituent knockout mouse programme (Duncan et al. 2011, van Dijk et al. 2014). (KOMP) of the Jackson Laboratory has recently Detailed phenotyping includes skeletal analysis commenced its own skeletal phenotyping project that during prenatal and postnatal development, at peak bone involves rapid micro-CT and automated bone and joint mass and in adulthood. Juvenile analysis establishes the cartilage histology (http://bonebase.org/). This screen role of gene in skeletal development and growth, whereas focuses on detecting evidence of variations in skeletal adult analysis establishes its role in skeletal maintenance cellular function. Histomorphometry is performed by a and repair. Gene expression pattern of the deleted gene recently innovated high-throughput process that involves can be investigated by LacZ staining, taking advantage computer-automated signal detection for the particular of the reporter gene included in the knockout-first gene cell type-specific stains. Data are accrued by automated targeting cassette. Temporal expression can be established analysis that calculates the percentage of the bone surface at different time points in embryonic and adult mice. containing the light signal from each stain, thereby To determine the cellular basis of the abnormal suggesting the pattern of disruption of cellular activity in phenotype, static and dynamic histomorphometry can the trabecular bone of the femur and vertebra that may be performed to identify abnormalities of osteoclastic account for the architectural observations seen in micro-CT bone resorption and osteoblastic bone formation (Bassett (Hong et al. 2012). Besides phenotyping inbred lines from et al. 2010). Primary chondrocytes, osteoblasts and the IKMC mutant mouse repository (Yoshiki & Moriwaki osteoclasts, cultures from wild-type and mutant mice 2006), the Bonebase phenotyping project is phenotyping can be undertaken to determine the consequences of mouse lines from the ‘Collaborative Cross’ project, which gene deletion on cell proliferation, differentiation and has created hybrids from eight founder inbred strains in function (Bassett et al. 2008). If these studies suggest that

Endocrinology order to perform genetic mapping studies to identify the the phenotype is a consequence of a defect in a specific of QTLs that contribute to complex traits and diseases (Bogue bone cell lineage, conditional deletion of the gene in et al. 2015). Similarly, Bonebase is also studying ‘diversity the specific cell type may be used to determine if the outbred’ lines created to produce a genetic resource to phenotype of the global knockout is recapitulated. By Journal facilitate high-resolution mapping of the effects of allelic crossing the knockout mouse line with an Flp deleter heterozygosity that replicates the complexity of the strain, the gene-trap knockout is reversed and a loxP- human population (Svenson et al. 2012). flanked (‘floxed’) allele is generatedFig. 2 ( ). Subsequently, floxed mice can be crossed with bone cell lineage-specific Cre strains (Murray et al. 2012) (Table 2). Current OBCD project goals To determine the molecular basis of the skeletal phenotype, the effect of the gene knockout on global The OBCD project is currently funded by a Wellcome Trust gene expression can be examined by whole-genome Strategic Award to undertake skeletal phenotyping of all microarray analysis of RNA extracted from bones of the knockout mouse lines generated at the Sanger Institute. knockout mice or from cell cultures. Cluster analysis is Results are available at the OBCD website and also performed to identify patterns of gene expression that uploaded to the IMPC mouse portal. Although the IMPC suggest involvement of specific signalling pathways. parent project is powered robustly to assess and catalogue In addition, whole transcriptome analysis can be the unknown pleiotropic effects of gene deletion, the performed by RNA sequencing with next-generation OBCD screen is designed for rapid-throughput hypothesis sequencing techniques. Bioinformatics analysis can generation. Once extreme phenotypes are detected, they determine the effect of the gene deletion on complex can be selected for additional in-depth analysis. gene networks including alternative splice variants and regulatory elements such as non-coding RNAs (Vikman et al. 2014). Detailed analysis of extreme phenotypes Although the rapid-throughput skeletal phenotyping Knockout mice with extreme skeletal phenotypes are screen is a powerful technique to identify new gene considered for additional detailed analysis and the that regulates bone mineralisation and strength, it has

Figure 5

Endocrinology Flow chart outlining selection of knockout mouse

of lines for further study and analysis.

limitations related to (i) the use of mice as a model system, can be investigated in large well-characterised human

Journal (ii) analysis of only knockout animals and (iii) the specific populations such as those included in the GEnetic skeletal phenotyping methods selected. Factors for OSteoporosis Consortium (GeFOS) (http:// There are well-described differences in bone www.gefos.org/). structure, bone remodelling and hormonal changes The IKMC/IMPC knockout strategy aims to determine between humans and mice. Despite these differences, the physiological role of genes by identifying the key molecules that regulate bone and cartilage have been pathophysiological consequences of loss-of-function but shown to have the same functions in mice and humans will not identify phenotypes associated with gain-of- and many inherited skeletal disorders are recapitulated function mutations, epigenetic modifications or other in genetically modified mice. In addition, although environmental risk factors. Nevertheless, if a significant there is no mouse menopause, the consequences of skeletal phenotype is detected in knockout animals, the oestrogen deficiency can still be studied in mice using the consequences of gain-of-function mutation could then be ovariectomy provocation model. Furthermore, if extreme studied by using CRISPR-Cas-based targeted gene-editing skeletal phenotypes are identified in mouse knockout techniques to generate mouse models (Gaj et al. 2013, lines, genetic variation in the homologous human genes Sander & Joung 2014).

Table 2 Examples of cell-specific promoter-driven Cre recombinases in available transgenic mouse lines

Mouse lines expressing Cre-recombinase Equivalent inducible Cell type with cell-specific promoter Cre-recombinase References Osteoblasts OC-Cre, Col1a1-Cre Col1-CreERT2 (Kim et al. 2004; Nakanishi et al. 2008; Zhang et al. 2002) Osteoclasts Ctsk-Cre, LysM-Cre Ctsk-CreERT2 (Chiu et al. 2004; Sanchez-Fernandez et al. 2012) Osteocytes Dmp1-Cre Dmp1-CreERT2 (Lu et al. 2007; Powell et al. 2011) Chondrocytes Col2-Cre Agc1-CreERT2 (Henry et al. 2009; Yoon et al. 2005)

Robust, sensitive and specific skeletal phenotyping Bassett JH, Williams AJ, Murphy E, Boyde A, Howell PG, Swinhoe R, screening requires a combination of complementary, Archanco M, Flamant F, Samarut J, Costagliola S, et al. 2008 A lack of thyroid hormones rather than excess thyrotropin causes abnormal rapid-throughput methodologies but can never be skeletal development in hypothyroidism. Molecular Endocrinology 22 exhaustive. However, once an extreme phenotype has 501–512. (doi:10.1210/me.2007-0221) been identified and prioritised for further detailed Bassett JH, Boyde A, Howell PG, Bassett RH, Galliford TM, Archanco M, Evans H, Lawson MA, Croucher P, St Germain DL, et al. 2010 Optimal analysis, other important determinants of bone bone strength and mineralization requires the type 2 iodothyronine strength such as tissue mineralisation and bone deiodinase in osteoblasts. PNAS 107 7604–7609. (doi:10.1073/ geometry can be determined by quantitative pnas.0911346107) Bassett JH, Gogakos A, White JK, Evans H, Jacques RM, van der Spek AH, backscattered electron scanning electron microscopy Ramirez-Solis R, Ryder E, Sunter D, Boyde A, et al. 2012a (Bassett et al. 2010) and statistical shape modelling Rapid-throughput skeletal phenotyping of 100 knockout mice (Yang et al. 2006), respectively. identifies 9 new genes that determine bone strength.PLoS Genetics 8 e1002858. (doi:10.1371/journal.pgen.1002858) In conclusion, systematic, rapid-throughput skeletal Bassett JH, van der Spek A, Gogakos A & Williams GR 2012b Quantitative phenotyping of genetically modified mice is an exciting X-ray imaging of rodent bone by Faxitron. Methods in Molecular new approach that complements human population Biology 816 499–506. (doi:10.1007/978-1-61779-415-5_29) Bogue MA, Churchill GA & Chesler EJ 2015 Collaborative cross and studies of complex polygenic disorders and has the diversity outbred data resources in the mouse phenome database. potential to identify important new signalling pathways Mammalian Genome 26 511–520. (doi:10.1007/s00335-015-9595-6) involved in the pathogenesis of skeletal disease. Bone HG, Dempster DW, Eisman JA, Greenspan SL, McClung MR, Nakamura T, Papapoulos S, Shih WJ, Rybak-Feiglin A, Santora AC, et al. 2015 Odanacatib for the treatment of postmenopausal osteoporosis: development history and design and participant characteristics of LOFT, the Long-Term Odanacatib Fracture Trial. Osteoporosis Declaration of interest International 26 699–712. (doi:10.1007/s00198-014-2944-6) The authors declare that there is no conflict of interest that could be Bouxsein ML, Boyd SK, Christiansen BA, Guldberg RE, Jepsen KJ & perceived as prejudicing the impartiality of this review. Muller R 2010 Guidelines for assessment of bone microstructure in rodents using micro-computed tomography. Journal of Bone and Mineral Research 25 1468–1486. (doi:10.1002/jbmr.141) Boyden LM, Mao J, Belsky J, Mitzner L, Farhi A, Mitnick MA, Wu D, Insogna K & Lifton RP 2002 High bone density due to a mutation in Endocrinology Funding LDL-receptor-related protein 5. New England Journal of Medicine 346 of This work was supported by a Wellcome Trust Strategic Award 1513–1521. (doi:10.1056/NEJMoa013444) (grant number 101123) and a Wellcome Trust Project Grant Bradley A, Anastassiadis K, Ayadi A, Battey JF, Bell C, Birling MC, (grant number 094134). Bottomley J, Brown SD, Bürger A, Bult CJ, et al. 2012 The mammalian Journal gene function resource: the international knockout mouse consortium. Mammalian Genome 23 580–586. (doi:10.1007/s00335- 012-9422-2) References Brommage R, Liu J, Hansen GM, Kirkpatrick LL, Potter DG, Sands AT, Abrahamsen B, van Staa T, Ariely R, Olson M & Cooper C 2009 Excess Zambrowicz B, Powell DR & Vogel P 2014 High-throughput screening mortality following hip fracture: a systematic epidemiological review. of mouse gene knockouts identifies established and novel skeletal Osteoporosis International 20 1633–1650. (doi:10.1007/s00198-009- phenotypes. Bone Research 2 14034. (doi:10.1038/boneres.2014.34) 0920-3) Brown SD & Moore MW 2012 The International Mouse Phenotyping Abrahamsen B, Grove EL & Vestergaard P 2014 Nationwide registry- Consortium: past and future perspectives on mouse phenotyping. based analysis of cardiovascular risk factors and adverse outcomes in Mammalian Genome 23 632–640. (doi:10.1007/s00335-012-9427-x) patients treated with strontium ranelate. Osteoporosis International 25 Brunkow ME, Gardner JC, Van Ness J, Paeper BW, Kovacevich BR, 757–762. (doi:10.1007/s00198-013-2469-4) Proll S, Skonier JE, Zhao L, Sabo PJ, Fu Y, et al. 2001 Bone dysplasia Alonso N & Ralston SH 2014 Unveiling the mysteries of the genetics sclerosteosis results from loss of the SOST gene product, a novel of osteoporosis. Journal of Endocrinological Investigation 37 925–934. cystine knot-containing protein. American Journal of Human Genetics (doi:10.1007/s40618-014-0149-7) 68 577–589. (doi:10.1086/318811) Baldridge D, Schwarze U, Morello R, Lennington J, Bertin TK, Pace JM, Cantor RM, Lange K & Sinsheimer JS 2010 Prioritizing GWAS Pepin MG, Weis M, Eyre DR, Walsh J, et al. 2008 CRTAP and LEPRE1 results: a review of statistical methods and recommendations for mutations in recessive osteogenesis imperfecta. Human Mutation 29 their application. American Journal of Human Genetics 86 6–22. 1435–1442. (doi:10.1002/humu.20799) (doi:10.1016/j.ajhg.2009.11.017) Balemans W, Ebeling M, Patel N, Van Hul E, Olson P, Dioszegi M, Chiu WS, McManus JF, Notini AJ, Cassady AI, Zajac JD & Davey RA 2004 Lacza C, Wuyts W, Van Den Ende J, Willems P, et al. 2001 Increased Transgenic mice that express Cre recombinase in osteoclasts. Genesis bone density in sclerosteosis is due to the deficiency of a novel 39 178–185. (doi:10.1002/gene.20041) secreted protein (SOST). Human Molecular Genetics 10 537–543. Choi YA, Lim J, Kim KM, Acharya B, Cho JY, Bae YC, Shin HI, Kim SY & (doi:10.1093/hmg/10.5.537) Park EK 2010 Secretome analysis of human BMSCs and identification Barrett-Connor E, Mosca L, Collins P, Geiger MJ, Grady D, Kornitzer M, of SMOC1 as an important ECM protein in osteoblast differentiation. McNabb MA & Wenger NK 2006 Effects of raloxifene on Journal of Proteome Research 9 2946–2956. (doi:10.1021/pr901110q) cardiovascular events and breast cancer in postmenopausal women. Chong B, Hegde M, Fawkner M, Simonet S, Cassinelli H, Coker M, New England Journal of Medicine 355 125–137. (doi:10.1056/ Kanis J, Seidel J, Tau C, Tuysuz B, et al. 2003 Idiopathic NEJMoa062462) hyperphosphatasia and TNFRSF11B mutations: relationships between

phenotype and genotype. Journal of Bone and Mineral Research 18 Goldring SR 2015 The osteocyte: key player in regulating bone turnover. 2095–2104. (doi:10.1359/jbmr.2003.18.12.2095) RMD Open 1 e000049. (doi:10.1136/rmdopen-2015-000049) Collins FS, Rossant J & Wurst W 2007 A mouse for all reasons. Cell 128 Gong Y, Slee RB, Fukai N, Rawadi G, Roman-Roman S, Reginato AM, 9–13. (doi:10.1016/j.cell.2006.12.018) Wang H, Cundy T, Glorieux FH, Lev D, et al. 2001 LDL receptor- Cooper DN 2010 Functional intronic polymorphisms: buried treasure related protein 5 (LRP5) affects bone accrual and eye development. awaiting discovery within our genes. Human Genomics 5 284–288. Cell 107 513–523. (doi:10.1016/S0092-8674(01)00571-2) (doi:10.1186/1479-7364-4-5-284) Gueguen R, Jouanny P, Guillemin F, Kuntz C, Pourel J & Siest G 1995 Coudert AE, de Vernejoul MC, Muraca M & Del Fattore A 2015 Segregation analysis and variance components analysis of bone Osteopetrosis and its relevance for the discovery of new functions mineral density in healthy families. Journal of Bone and Mineral associated with the skeleton. International Journal of Endocrinology Research 10 2017–2022. (doi:10.1002/jbmr.5650101223) 2015 372156. (doi:10.1155/2015/265151) Guerrini MM, Sobacchi C, Cassani B, Abinun M, Kilic SS, Pangrazio A, Cox RD & Church CD 2011 Mouse models and the interpretation of Moratto D, Mazzolari E, Clayton-Smith J, Orchard P, et al. 2008 human GWAS in type 2 diabetes and obesity. Disease Models & Human osteoclast-poor osteopetrosis with hypogammaglobulinemia Mechanisms 4 155–164. (doi:10.1242/dmm.000414) due to TNFRSF11A (RANK) mutations. American Journal of Human de Angelis MH, Nicholson G, Selloum M, White JK, Morgan H, Genetics 83 64–76. (doi:10.1016/j.ajhg.2008.06.015) Ramirez-Solis R, Sorg T, Wells S, Fuchs H, Fray M, et al. 2015 Analysis Hangauer MJ, Vaughn IW & McManus MT 2013 Pervasive transcription of mammalian gene function through broad-based phenotypic of the human genome produces thousands of previously unidentified screens across a consortium of mouse clinics. Nature Genetics 47 long intergenic noncoding RNAs. PLoS Genetics 9 e1003569. 969–978. (doi:10.1038/ng.3360) (doi:10.1371/journal.pgen.1003569) Delany AM, Amling M, Priemel M, Howe C, Baron R & Canalis E 2000 Heino TJ, Hentunen TA & Vaananen HK 2002 Osteocytes inhibit Osteopenia and decreased bone formation in osteonectin-deficient osteoclastic bone resorption through transforming growth mice. Journal of Clinical Investigation 105 915–923. (doi:10.1172/ factor-beta: enhancement by estrogen. Journal of Cellular Biochemistry JCI7039) 85 185–197. (doi:10.1002/jcb.10109) Deng FY, Lei SF, Chen XD, Tan LJ, Zhu XZ & Deng HW 2011a An Henry SP, Jang CW, Deng JM, Zhang Z, Behringer RR & de integrative study ascertained SOD2 as a susceptibility gene for Crombrugghe B 2009 Generation of aggrecan-CreERT2 knockin mice osteoporosis in Chinese. Journal of Bone and Mineral Research 26 for inducible Cre activity in adult cartilage. Genesis 47 805–814. 2695–2701. (doi:10.1002/jbmr.471) (doi:10.1002/dvg.20564) Deng FY, Lei SF, Zhang Y, Zhang YL, Zheng YP, Zhang LS, Pan R, Wang L, Hernlund E, Svedbom A, Ivergard M, Compston J, Cooper C, Stenmark J, Tian Q, Shen H, et al. 2011b Peripheral blood monocyte-expressed McCloskey EV, Jonsson B & Kanis JA 2013 Osteoporosis in the ANXA2 gene is involved in pathogenesis of osteoporosis in humans. European Union: medical management, epidemiology and economic Molecular & Cellular Proteomics 10 M111.011700. (doi:10.1074/mcp. burden. A report prepared in collaboration with the International m111.011700) Osteoporosis Foundation (IOF) and the European Federation of Dolan ME, Baldarelli RM, Bello SM, Ni L, McAndrews MS, Bult CJ, Pharmaceutical Industry Associations (EFPIA). Archives of Osteoporosis Endocrinology Kadin JA, Richardson JE, Ringwald M, Eppig JT, et al. 2015 8 136. (doi:10.1007/s11657-013-0136-1) of Orthology for comparative genomics in the mouse genome Holmen SL, Giambernardi TA, Zylstra CR, Buckner-Berghuis BD, database. Mammalian Genome 26 305–313. (doi:10.1007/s00335- Resau JH, Hess JF, Glatt V, Bouxsein ML, Ai M, Warman ML, et al. 015-9588-5) 2004 Decreased BMD and limb deformities in mice carrying Journal Duncan EL, Danoy P, Kemp JP, Leo PJ, McCloskey E, Nicholson GC, mutations in both Lrp5 and Lrp6. Journal of Bone and Mineral Research Eastell R, Prince RL, Eisman JA, Jones G, et al. 2011 Genome-wide 19 2033–2040. (doi:10.1359/jbmr.040907) association study using extreme truncate selection identifies novel Holroyd C, Harvey N, Dennison E & Cooper C 2012 Epigenetic genes affecting bone mineral density and fracture risk. PLoS Genetics influences in the developmental origins of osteoporosis.Osteoporosis 7 e1001372. (doi:10.1371/journal.pgen.1001372) International 23 401–410. (doi:10.1007/s00198-011-1671-5) Esapa CT, Bassett JH, Evans H, Croucher PI, Williams GR & Thakker RV Hong SH, Jiang X, Chen L, Josh P, Shin DG & Rowe D 2012 2012 Bone mineral content and density. Current Protocols in Mouse Computer-automated static, dynamic and cellular bone Biology 2 365–400. (doi:10.1002/9780470942390.mo120124) histomorphometry. Journal of Tissue Science & Engineering 5 Estrada K, Styrkarsdottir U, Evangelou E, Hsu YH, Duncan EL, Ntzani EE, (Supplement 1) 004. (doi:10.4172/2157-7552.S1-004) Oei L, Albagha OM, Amin N, Kemp JP, et al. 2012 Genome-wide International HapMap Consortium 2005 A haplotype map of the human meta-analysis identifies 56 bone mineral density loci and reveals genome. Nature 437 1299–1320. (doi:10.1038/nature04226) 14 loci associated with risk of fracture. Nature Genetics 44 491–501. Ioannidis JP, Ng MY, Sham PC, Zintzaras E, Lewis CM, Deng HW, (doi:10.1038/ng.2249) Econs MJ, Karasik D, Devoto M, Kammerer CM, et al. 2007 Frattini A, Orchard PJ, Sobacchi C, Giliani S, Abinun M, Mattsson JP, Meta-analysis of genome-wide scans provides evidence for sex- and Keeling DJ, Andersson AK, Wallbrandt P, Zecca L, et al. 2000 Defects site-specific regulation of bone mass.Journal of Bone and Mineral in TCIRG1 subunit of the vacuolar proton pump are responsible for a Research 22 173–183. (doi:10.1359/jbmr.060806) subset of human autosomal recessive osteopetrosis. Nature Genetics 25 Johnell O & Kanis JA 2006 An estimate of the worldwide prevalence 343–346. (doi:10.1038/77131) and disability associated with osteoporotic fractures. Osteoporosis Frazer KA, Murray SS, Schork NJ & Topol EJ 2009 Human genetic International 17 1726–1733. (doi:10.1007/s00198-006-0172-4) variation and its contribution to complex traits. Nature Reviews Johnell O, Kanis JA, Oden A, Johansson H, De Laet C, Delmas P, Genetics 10 241–251. (doi:10.1038/nrg2554) Eisman JA, Fujiwara S, Kroger H, Mellstrom D, et al. 2005 Predictive Gaj T, Gersbach CA & Barbas CF 3rd 2013 ZFN, TALEN, and value of BMD for hip and other fractures. Journal of Bone and Mineral CRISPR/Cas-based methods for genome engineering. Trends in Research 20 1185–1194. (doi:10.1359/JBMR.050304) Biotechnology 31 397–405. (doi:10.1016/j.tibtech.2013.04.004) Johnson ML, Gong G, Kimberling W, Reckér SM, Kimmel DB & Glass DA 2nd, Bialek P, Ahn JD, Starbuck M, Patel MS, Clevers H, Recker RB 1997 Linkage of a gene causing high bone mass to human Taketo MM, Long F, McMahon AP, Lang RA & Karsenty G 2005 chromosome 11 (11q12-13). American Journal of Human Genetics 60 Canonical Wnt signaling in differentiated osteoblasts controls 1326–1332. (doi:10.1086/515470) osteoclast differentiation. Developmental Cell 8 751–764. Karp NA, Meehan TF, Morgan H, Mason JC, Blake A, Kurbatova N, (doi:10.1016/j.devcel.2005.02.017) Smedley D, Jacobsen J, Mott RF, Iyer V, et al. 2015 Applying the

ARRIVE guidelines to an in vivo database. PLoS Biology 13 e1002151. Nakanishi R, Akiyama H, Kimura H, Otsuki B, Shimizu M, Tsuboyama T (doi:10.1371/journal.pbio.1002151) & Nakamura T 2008 Osteoblast-targeted expression of Sfrp4 in mice Keen RW, Hart DJ, Arden NK, Doyle DV & Spector TD 1999 Family results in low bone mass. Journal of Bone and Mineral Research 23 history of appendicular fracture and risk of osteoporosis: a 271–277. (doi:10.1359/jbmr.071007) population-based study. Osteoporosis International 10 161–166. Nakashima T, Hayashi M, Fukunaga T, Kurata K, Oh-Hora M, Feng JQ, (doi:10.1007/s001980050211) Bonewald LF, Kodama T, Wutz A, Wagner EF, et al. 2011 Evidence for Kelly PJ, Nguyen T, Hopper J, Pocock N, Sambrook P & Eisman J 1993 osteocyte regulation of bone homeostasis through RANKL expression. Changes in axial bone density with age: a twin study. Journal of Bone Nature Medicine 17 1231–1234. (doi:10.1038/nm.2452) and Mineral Research 8 11–17. (doi:10.1002/jbmr.5650080103) Pangrazio A, Poliani PL, Megarbane A, Lefranc G, Lanino E, Di Rocco M, Kennedy OD, Herman BC, Laudier DM, Majeska RJ, Sun HB & Rucci F, Lucchini F, Ravanini M, Facchetti F, et al. 2006 Mutations in Schaffler MB 2012 Activation of resorption in fatigue-loaded bone OSTM1 (grey lethal) define a particularly severe form of autosomal involves both apoptosis and active pro-osteoclastogenic signaling by recessive osteopetrosis with neural involvement. Journal of Bone and distinct osteocyte populations. Bone 50 1115–1122. (doi:10.1016/ Mineral Research 21 1098–1105. (doi:10.1359/jbmr.060403) j.bone.2012.01.025) Powell WF Jr, Barry KJ, Tulum I, Kobayashi T, Harris SE, Bringhurst FR Kennel KA & Drake MT 2009 Adverse effects of bisphosphonates: & Pajevic PD 2011 Targeted ablation of the PTH/PTHrP receptor in implications for osteoporosis management. Mayo Clinic Proceedings 84 osteocytes impairs bone structure and homeostatic calcemic responses. 632–637. (doi:10.1016/S0025-6196(11)60752-0) Journal of Endocrinology 209 21–32. (doi:10.1530/joe-10-0308 ) Kim JE, Nakashima K & de Crombrugghe B 2004 Transgenic mice Raggatt LJ & Partridge NC 2010 Cellular and molecular mechanisms of expressing a ligand-inducible cre recombinase in osteoblasts and bone remodeling. Journal of Biological Chemistry 285 25103–25108. odontoblasts: a new tool to examine physiology and disease of (doi:10.1074/jbc.R109.041087) postnatal bone and tooth. American Journal of Pathology 165 Richards JB, Zheng HF & Spector TD 2012 Genetics of osteoporosis from 1875–1882. (doi:10.1016/S0002-9440(10)63240-3) genome-wide association studies: advances and challenges. Nature Kornak U, Kasper D, Bosl MR, Kaiser E, Schweizer M, Schulz A, Reviews Genetics 13 576–588. (doi:10.1038/nrg3228) Friedrich W, Delling G & Jentsch TJ 2001 Loss of the ClC-7 chloride Robling AG, Niziolek PJ, Baldridge LA, Condon KW, Allen MR, Alam I, channel leads to osteopetrosis in mice and man. Cell 104 205–215. Mantila SM, Gluhak-Heinrich J, Bellido TM, Harris SE, et al. 2008 (doi:10.1016/S0092-8674(01)00206-9) Mechanical stimulation of bone in vivo reduces osteocyte expression Koscielny G, Yaikhom G, Iyer V, Meehan TF, Morgan H, Atienza-Herrero J, of Sost/sclerostin. Journal of Biological Chemistry 283 5866–5875. Blake A, Chen CK, Easty R, Di Fenza A, et al. 2014 The International (doi:10.1074/jbc.M705092200) Mouse Phenotyping Consortium Web Portal, a unified point of Rousseeuw P & Van Zomeren B 1990 Unmasking multivariate outliers access for knockout mice and related phenotyping data. Nucleic Acids and leverage points. Journal of the American Statistical Association 85 Research 42 D802–D809. (doi:10.1093/nar/gkt977) 633–639. (doi:10.1080/01621459.1990.10474920) Lacey DL, Boyle WJ, Simonet WS, Kostenuik PJ, Dougall WC, Sullivan JK, Sanchez-Fernandez MA, Sbacchi S, Correa-Tapia M, Naumann R, San Martin J & Dansey R 2012 Bench to bedside: elucidation of the Klemm J, Chambon P, Al-Robaiy S, Blessing M & Hoflack B 2012 Endocrinology OPG-RANK-RANKL pathway and the development of denosumab. Transgenic mice for a tamoxifen-induced, conditional expression of of Nature Reviews. Drug Discovery 11 401–419. (doi:10.1038/nrd3705) the Cre recombinase in osteoclasts. PLoS ONE 7 e37592. (doi:10.1371/ Little RD, Carulli JP, Del Mastro RG, Dupuis J, Osborne M, Folz C, journal.pone.0037592) Manning SP, Swain PM, Zhao SC, Eustace B, et al. 2002 A mutation Sander JD & Joung JK 2014 CRISPR-Cas systems for editing, regulating Journal in the LDL receptor-related protein 5 gene results in the autosomal and targeting genomes. Nature Biotechnology 32 347–355. dominant high-bone-mass trait. American Journal of Human Genetics (doi:10.1038/nbt.2842) 2002 70 11–19. (doi:10.1086/338450) Sattui SE & Saag KG 2014 Fracture mortality: associations with Lu Y, Xie Y, Zhang S, Dusevich V, Bonewald LF & Feng JQ 2007 epidemiology and osteoporosis treatment. Nature Reviews DMP1-targeted Cre expression in odontoblasts and osteocytes. Journal Endocrinology 10 592–602. (doi:10.1038/nrendo.2014.125) of Dental Research 86 320–325. (doi:10.1177/154405910708600404) Schofield PN, Hoehndorf R & Gkoutos GV 2012 Mouse genetic and Makras P, Delaroudis S & Anastasilakis AD 2015 Novel therapies for phenotypic resources for human genetics. Human Mutation 33 osteoporosis. Metabolism 64 1199–1214. (doi:10.1016/j.metabol. 826–836. (doi:10.1002/humu.22077) 2015.07.011) Seeman E, Hopper JL, Bach LA, Cooper ME, Parkinson E, McKay J Manolio TA 2010 Genomewide association studies and assessment of & Jerums G 1989 Reduced bone mass in daughters of women the risk of disease. New England Journal of Medicine 363 166–176. with osteoporosis. New England Journal of Medicine 320 554–558. (doi:10.1056/NEJMra0905980) (doi:10.1056/NEJM198903023200903) Marshall D, Johnell O & Wedel H 1996 Meta-analysis of how well Shah AD, Shoback D & Lewiecki EM 2015 Sclerostin inhibition: a novel measures of bone mineral density predict occurrence of osteoporotic therapeutic approach in the treatment of osteoporosis. International fractures. BMJ 312 1254–1259. (doi:10.1136/bmj.312.7041.1254) Journal of Women’s Health 7 565–580. (doi:10.2147/ijwh.s73244) Mitchell BD & Streeten EA 2013 Clinical impact of recent genetic Sim I-W & Ebeling PR 2015 Romosozumab/CDP7851 for the treatment of discoveries in osteoporosis. Application of Clinical Genetics 6 75–85. osteoporosis. Expert Review of Endocrinology & Metabolism 10 471–481. (doi:10.2147/TACG.S52047) (doi:10.1586/17446651.2015.1079482) Mitchell BD & Yerges-Armstrong LM 2011 The genetics of bone loss: Simonet WS, Lacey DL, Dunstan CR, Kelley M, Chang MS, challenges and prospects. Journal of Clinical Endocrinology and Luthy R, Nguyen HQ, Wooden S, Bennett L, Boone T, et al. Metabolism 96 1258–1268. (doi:10.1210/jc.2010-2865) 1997 Osteoprotegerin: a novel secreted protein involved in the Montague CT, Farooqi IS, Whitehead JP, Soos MA, Rau H, Wareham NJ, regulation of bone density. Cell 89 309–319. (doi:10.1016/S0092- Sewter CP, Digby JE, Mohammed SN, Hurst JA, et al. 1997 Congenital 8674(00)80209-3) leptin deficiency is associated with severe early-onset obesity in Skarnes WC, Rosen B, West AP, Koutsourakis M, Bushell W, Iyer V, humans. Nature 387 903–908. (doi:10.1038/43185) Mujica AO, Thomas M, Harrow J, Cox T, et al. 2011 A conditional Murray SA, Eppig JT, Smedley D, Simpson EM & Rosenthal N knockout resource for the genome-wide study of mouse gene 2012 Beyond knockouts: cre resources for conditional function. Nature 474 337–342. (doi:10.1038/nature10163) mutagenesis. Mammalian Genome 23 587–599. (doi:10.1007/ Sobacchi C, Frattini A, Guerrini MM, Abinun M, Pangrazio A, Susani L, s00335-012-9430-2) Bredius R, Mancini G, Cant A, Bishop N, et al. 2007 Osteoclast-poor

human osteopetrosis due to mutations in the gene encoding RANKL. Waterston RH, Lindblad-Toh K, Birney E, Rogers J, Abril JF, Agarwal P, Nature Genetics 39 960–962. (doi:10.1038/ng2076) Agarwala R, Ainscough R, Alexandersson M, An P, et al. 2002 Initial Svenson KL, Gatti DM, Valdar W, Welsh CE, Cheng R, Chesler EJ, sequencing and comparative analysis of the mouse genome. Nature Palmer AA, McMillan L & Churchill GA 2012 High-resolution genetic 420 520–562. (doi:10.1038/nature01262) mapping using the mouse diversity outbred population. Genetics 190 White JK, Gerdin AK, Karp NA, Ryder E, Buljan M, Bussell JN, Salisbury J, 437–447. (doi:10.1534/genetics.111.132597) Clare S, Ingham NJ, Podrini C, et al. 2013 Genome-wide generation Sykes B, Ogilvie D, Wordsworth P, Anderson & Jones N 1986 Osteogenesis and systematic phenotyping of knockout mice reveals new roles for imperfecta is linked to both type I collagen structural genes. Lancet 2 many genes. Cell 154 452–464. (doi:10.1016/j.cell.2013.06.022) 69–72. (doi:10.1016/S0140-6736(86)91609-0) Wu S, Liu Y, Zhang L, Han Y, Lin Y & Deng HW 2013 Genome-wide Testa G, Schaft J, van der Hoeven F, Glaser S, Anastassiadis K, Zhang Y, approaches for identifying genetic risk factors for osteoporosis. Hermann T, Stremmel W & Stewart AF 2004 A reliable lacZ expression Genome Medicine 5 44. (doi:10.1186/gm448) reporter cassette for multipurpose, knockout-first alleles.Genesis 38 Yamagata K, Oda N, Kaisaki PJ, Menzel S, Furuta H, Vaxillaire M, 151–158. (doi:10.1002/gene.20012) Southam L, Cox RD, Lathrop GM, Boriraj VV, et al. 1996 Mutations Trost Z, Trebse R, Prezelj J, Komadina R, Logar DB & Marc J 2010 in the hepatocyte nuclear factor-1alpha gene in maturity- A microarray based identification of osteoporosis-related genes in onset diabetes of the young (MODY3). Nature 384 455–458. primary culture of human osteoblasts. Bone 46 72–80. (doi:10.1016/ (doi:10.1038/384455a0) j.bone.2009.09.015) Yang Y, Bull A, Rueckert D & Hill A 2006 3D statistical shape modeling Tu X, Rhee Y, Condon KW, Bivi N, Allen MR, Dwyer D, Stolina M, of long bones. In Biomedical Image Registration, pp 306–314. Eds JPW Turner CH, Robling AG, Plotkin LI, et al. 2012 Sost downregulation Pluim, B Likar & FA Gerritsen. Berlin, Germany: Springer-Verlag. and local Wnt signaling are required for the osteogenic response to (doi:10.1007/11784012_37) mechanical loading. Bone 50 209–217. (doi:10.1016/j.bone. Yoon BS, Ovchinnikov DA, Yoshii I, Mishina Y, Behringer RR & 2011.10.025) Lyons KM 2005 Bmpr1a and Bmpr1b have overlapping functions van Dijk FS, Nesbitt IM, Zwikstra EH, Nikkels PG, Piersma SR, and are essential for chondrogenesis in vivo. PNAS 102 5062–5067. Fratantoni SA, Jimenez CR, Huizer M, Morsman AC, Cobben JM, (doi:10.1073/pnas.0500031102) et al. 2009 PPIB mutations cause severe osteogenesis imperfecta. Yoshiki A & Moriwaki K 2006 Mouse phenome research: implications of American Journal of Human Genetics 85 521–527. (doi:10.1016/ genetic background. ILAR Journal 47 94–102. (doi:10.1093/ilar.47.2.94) j.ajhg.2009.09.001) Zhang M, Xuan S, Bouxsein ML, von Stechow D, Akeno N, Faugere MC, van Dijk FS, Zillikens MC, Micha D, Riessland M, Marcelis CL, Malluche H, Zhao G, Rosen CJ, Efstratiadis A, et al. 2002 de Die-Smulders CE, Milbradt J, Franken AA, Harsevoort AJ, Osteoblast-specific knockout of the insulin-like growth factor (IGF) Lichtenbelt KD, et al. 2013 PLS3 mutations in X-linked osteoporosis receptor gene reveals an essential role of IGF signaling in bone matrix with fractures. New England Journal of Medicine 369 1529–1536. mineralization. Journal of Biological Chemistry 277 44005–44012. (doi:10.1056/NEJMoa1308223) (doi:10.1074/jbc.M208265200) van Dijk EL, Auger H, Jaszczyszyn Y & Thermes C 2014 Ten years of Zhang JG, Tan LJ, Xu C, He H, Tian Q, Zhou Y, Qiu C, Chen XD & Endocrinology next-generation sequencing technology. Trends in Genetics 30 Deng HW 2015 Integrative analysis of transcriptomic and epigenomic of 418–426. (doi:10.1016/j.tig.2014.07.001) data to reveal regulation patterns for BMD variation. PLoS ONE 10 Vikman P, Fadista J & Oskolkov N 2014 RNA sequencing: current e0138524. (doi:10.1371/journal.pone.0138524) and prospective uses in metabolic research. Journal of Molecular Zheng HF, Forgetta V, Hsu YH, Estrada K, Rosello-Diez A, Leo PJ, Journal Endocrinology 53 R93–R101. (doi:10.1530/JME-14-0170) Dahia CL, Park-Min KH, Tobias JH, Kooperberg C, et al. 2015 Visscher PM, Brown MA, McCarthy MI & Yang J 2012 Five years of GWAS Whole-genome sequencing identifies EN1 as a determinant of discovery. American Journal of Human Genetics 90 7–24. (doi:10.1016/ bone density and fracture. Nature 526 112–117. (doi:10.1038/ j.ajhg.2011.11.029) nature14878)

Received in final form 11 August 2016 Accepted 17 August 2016 Accepted Preprint published online 17 August 2016